Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

Subject: Re: [OMPI users] opal_os_dirpath_create: Error: Unable to create the, sub-directory
From: Ralph Castain (rhc_at_[hidden])
Date: 2014-02-03 14:49:02


Seems rather odd - is your /tmp by any chance network mounted?

On Feb 3, 2014, at 9:41 AM, Eric Chamberland <Eric.Chamberland_at_[hidden]> wrote:

> Hi,
>
> with OpenMPI 1.6.3 I have encountered this error which "randomly" appears:
>
> [compile:20089] opal_os_dirpath_create: Error: Unable to create the sub-directory (/tmp/openmpi-sessions-cmpbib_at_compile_0/55528/0) of (/tmp/openmpi-sessions-cmpbib_at_compile_0/55528/0/0), mkdir failed [1]
> [compile:20089] [[55528,0],0] ORTE_ERROR_LOG: Error in file util/session_dir.c at line 106
>
> (view full stderr attached)
>
> and also this mostly same one:
>
> [compile:22876] opal_os_dirpath_create: Error: Unable to create the sub-directory (/tmp/openmpi-sessions-cmpbib_at_compile_0/53197/0) of (/tmp/openmpi-sessions-cmpbib_at_compile_0/53197/0/0), mkdir failed [1]
> ...
>
> Looking deeper, I have found this in /tmp:
>
> ls -ladtr /tmp/openmpi-sessions-cmpbib\@compile_0/* |grep -v "drwx"
> -rw-r--r-- 1 cmpbib bib 93 Jan 31 06:47 /tmp/openmpi-sessions-cmpbib_at_compile_0/55528
> -rw-r--r-- 1 cmpbib bib 92 Jan 31 06:48 /tmp/openmpi-sessions-cmpbib_at_compile_0/41437
> -rw-r--r-- 1 cmpbib bib 93 Jan 31 07:01 /tmp/openmpi-sessions-cmpbib_at_compile_0/59324
> -rw-r--r-- 1 cmpbib bib 92 Jan 31 09:49 /tmp/openmpi-sessions-cmpbib_at_compile_0/53197
> -rw-r--r-- 1 cmpbib bib 93 Jan 31 11:10 /tmp/openmpi-sessions-cmpbib_at_compile_0/54532
> -rw-r--r-- 1 cmpbib bib 93 Jan 31 14:18 /tmp/openmpi-sessions-cmpbib_at_compile_0/36511
> -rw-r--r-- 1 cmpbib bib 93 Feb 1 18:50 /tmp/openmpi-sessions-cmpbib_at_compile_0/63980
>
>
> So there are some *files* in /tmp which are named like the directories which are tried to be created....
>
> The content of the file /tmp/openmpi-sessions-cmpbib_at_compile_0/55528 is:
>
> 4016963584.0;tcp://10.1.1.46:51427;tcp://132.203.7.103:51427;tcp://192.168.122.1:51427
> 31231
>
> which looks like the content of the file "contact.txt" which seems to appear in a successfully created directory. Also, the files have been created far before the executions which aborted...
>
> So, is this a bug in 1.6.3 and is there a "solution" for that?
> (I know I can cleanup the files, but I expect OpenMPI to not try to create a directory if a file with the same name exists...)
>
> Thanks,
>
> Eric
> <Cerr.DiffusionInst.2.13.txt><Cerr.mooney_3d_Clpen_mixte_64Hexa8_3proc.txt><ompi_info.txt>_______________________________________________
> users mailing list
> users_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/users